当前位置: X-MOL 学术Comput. Biol. Med. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Semi-supervised generative adversarial networks for the segmentation of the left ventricle in pediatric MRI.
Computers in Biology and Medicine ( IF 7.0 ) Pub Date : 2020-06-29 , DOI: 10.1016/j.compbiomed.2020.103884
Colin Decourt 1 , Luc Duong 2
Affiliation  

Segmentation of the left ventricle in magnetic resonance imaging (MRI) is important for assessing cardiac function. We present DT-GAN, a generative adversarial network (GAN) segmentation approach for the identification of the left ventricle in pediatric MRI. Segmentation of the left ventricle requires a large amount of annotated data; generating such data can be time-consuming and subject to observer variability. Additionally, it can be difficult to accomplish in a clinical setting. During the training of our GAN, we therefore introduce a semi-supervised semantic segmentation to reduce the number of images required for training, while maintaining a good segmentation accuracy. The GAN generator produces a segmentation label map and its discriminator outputs a confidence map, which gives the probability of a pixel coming from the label or from the generator. Moreover, we propose a new formulation of the GAN loss function based on distance transform and pixel-wise cross-entropy. This new loss function provides a better segmentation of boundary pixels, by favoring the correct classification of those pixels rather than focusing on pixels that are farther away from the boundary between anatomical structures. Our proposed method achieves a mean Hausdorff distance of 2.16 mm ± 0.42 mm (2.28 mm ± 0.21 mm for U-Net) and a Dice score of 0.88 ± 0.08 (0.91 ± 0.12 for U-Net) for the endocardium segmentation, using 50% of the annotated data. For the epicardium segmentation, we achieve a mean Hausdorff distance of 2.23 mm ± 0.35 mm (2.34 mm ± 0.39 mm for U-Net) and a Dice score of 0.93 mm ± 0.04 mm (0.89 ± 0.09 for U-Net). For the myocardium segmentation, we achieve a mean Hausdorff distance of 2.98 mm ± 0.43 mm (3.04 mm ± 0.27 mm for U-Net) and a Dice score of 0.79 mm ± 0.10 mm (0.74 ± 0.04 for U-Net). This new model could be very useful for the automatic analysis of cardiac MRI and for conducting large-scale studies based on MRI readings, with a limited amount of training data.



中文翻译:

小儿MRI左心室分割的半监督生成对抗网络。

磁共振成像(MRI)中左心室的分割对于评估心脏功能很重要。我们目前提出DT-GAN,一种生成的对抗网络(GAN)分割方法,用于在儿科MRI中识别左心室。左心室的分割需要大量的注释数据。生成此类数据可能很耗时,并且会因观察者的不同而异。另外,在临床环境中可能难以完成。因此,在我们的GAN训练过程中,我们引入了半监督语义分割,以减少训练所需的图像数量,同时保持良好的分割精度。GAN生成器会生成一个细分标签图,其鉴别器会输出一个置信度图,给出像素来自标签或生成器的概率。此外,我们提出了一种基于距离变换和逐像素交叉熵的GAN损失函数的新公式。通过支持对这些像素进行正确的分类,而不是专注于距离解剖结构之间的边界更远的像素,此新的损失函数可以更好地对边界像素进行分割。我们提出的方法实现的平均Hausdorff距离为2.16 mm 通过支持对这些像素进行正确的分类,而不是专注于远离解剖结构之间的边界的像素。我们提出的方法实现的平均Hausdorff距离为2.16 mm 通过支持对这些像素进行正确的分类,而不是专注于远离解剖结构之间的边界的像素。我们提出的方法实现的平均Hausdorff距离为2.16 mm± 0.42毫米(2.28毫米 ±使用50%的注释数据,对于心内膜分割,U-Net为0.21 mm)和Dice得分为0.88±0.08(U-Net为0.91±0.12)。对于心外膜分割,我们获得的平均Hausdorff距离为2.23 mm± 0.35毫米(2.34毫米 ± U-Net为0.39毫米),骰子得分为0.93毫米 ±0.04毫米(U-Net为0.89±0.09)。对于心肌分割,我们获得的平均Hausdorff距离为2.98 mm± 0.43毫米(3.04毫米 ± U-Net为0.27毫米),骰子得分为0.79毫米 ±0.10毫米(对于U-Net为0.74±0.04)。这种新模型对于心脏MRI的自动分析以及基于MRI读数进行有限数量的训练数据进行大规模研究非常有用。

更新日期:2020-07-03
down
wechat
bug